Accelerating non-Cartesian sense for large coil arrays: Application to motion compensation in multishot DWI

Multi-shot sequences are often combined with techniques such as parallel imaging to achieve high fidelity images. For non-Cartesian sampling schemes, the reconstruction of such data becomes extremely computationally intensive. The problem of motion compensated SENSE reconstruction of non-Cartesian diffusion weighted images falls into a similar setting, when phase corrections are involved. Here we propose a new pipeline for the fast reconstruction of such data. The large array of composite sensitivity functions are replaced by a low-dimensional set of virtual sensitivity functions using a principal component analysis, thus enabling the evaluation of very few Fourier transforms. The time consuming gridding steps are replaced by a more efficient multiplication in the k-space, enabling further simplifications. Significant acceleration of reconstruction time is shown to be achieved with the new scheme. The algorithm in the general setting can accelerate SENSE reconstruction for large coil arrays.